Agricultural crop product pattern detection using optical and radar images with extreme gradient boosting algorithm

In this study, the effect of Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 (Multispectral) data on classification and agricultural crop pattern detection was investigated. The study area covers an area of approximately 2200 km2 within the borders of & Ccedil;ukurova Plain. Within the scope of the study, agricultural crop pattern classification including corn, cotton, wheat, sunflower, watermelon, peanut and citrus trees, as well as second crop corn, soybean and cotton crops planted after wheat, was performed using the extreme gradient boosting (XGBoost) algorithm with multi -temporal Sentinel-1 and Sentinel-2 images of 2021. Parcels registered in the Farmer Registration System (FRS) were used as reference parcels in the study. Before using the FRS data as ground truth data, pre-editing and rule-based deletion processes were performed, and then erroneous and false declarations were eliminated. In the study, the overall accuracy of the classification using only Sentinel-1 data (VH, VV, VH/VV) was 72.3%, and the overall accuracy of the classification using only Sentinel-2 data (R, G, B, NIR, NDVI) was 87.2%, and the overall accuracy of the classification using Sentinel-1 and Sentinel-2 data together was 92.1%. When the classification study was analyzed on a crop basis, it was determined that the lowest accuracy belonged to the classes calculated only with Sentinel-1 data, while the highest accuracy rate belonged to the study in which Sentinel-1 and Sentinel-2 data were used together. Especially in the second crop with very close phenological periods, it was observed that the use of Sentinel-1 and Sentinel-2 data together increased the success rate considerably.